Bank Lending Channel of Monetary Transmission in Indonesia
-
Upload
njprastowo -
Category
Documents
-
view
368 -
download
1
description
Transcript of Bank Lending Channel of Monetary Transmission in Indonesia
Bank Lending Channel of
Monetary Transmission in Indonesia
Juda Agung* Rita Morena
Bambang Pramono Nugroho Joko Prastowo
Directorate of Economic Research and Monetary Policy
BANK INDONESIA
2001
_________________________
* The authors thank Hartadi A Sarwono, Perry Warjiyo, Sjamsul Arifin, Wibisono, and Sri
Liani Suselo for their consistent encouragement and thoughtful comments to this
study. Of course, any errors and ommisions are our responsibility. The findings and
conclusions in this study are those of the authors and do not necessarily represent the
views of Bank Indonesia.
1
Abstract
Using a battery of tests, the study investigates the existence of bank
lending channel of monetary transmission in Indonesia before and after crisis.
Aggregate evidence shows that a monetary policy is able to affect bank
lending with a lag due to ability of banks to insulate the decrease in deposits by
liquidating their securities holdings. Disaggregate evidence show that in the
aftermath of a monetary policy shock, there is a „flight to quality‟ of deposits
especially from private domestic banks to foreign banks and state banks and
„flight to quality‟ of bank lending from individuals to firms. Results of panel data
estimation demonstrate that effect of monetary policy is stronger for low capital
banks. Furthermore, survey on banks and firms support the econometric results.
We also found that that efficacy of a monetary policy, particularly a monetary
contraction, in influencing the bank lending is stronger in the period of post crisis
than prior to the crisis. This findings lends support indirectly existence of
asymmetric effect of a monetary policy: the stronger in the recession than in the
boom periods, stronger for low capital banks than low capital and for less
creditworthy borrowers.
JEL Classification: E44, E50
Key words: bank lending channel, bank portfolio behaviour, monetary
transmission
2
1. Introduction
Monetary policy in Indonesia is recently being faced with the most
challenging time. The monetary policy to control high pressure of inflation
and smooth out volatility of Rupiah exchange rate has been constrained
by banks and firms‟ financial restructuring process. Under such
circumstances when disintermediation of banking system takes place,
efficacy of monetary policy has been declining and sometimes the policy
is seen by many observers as costly.
The problems have been aggravated by the uncertainty regarding
the way in which the monetary policy affects the real economy in the
aftermath of the crisis. Attempts have been made to understanding the
monetary policy implications of current the banking crisis. A study by
Agung, et al (2001) to understand the existence of financial
disintermediation in the banking sector in the aftermath of the crisis and its
monetary policy implications is one of the attempts. However, a broader
agenda to understand the whole picture of monetary transmission
mechanism needs to be done. This agenda is of paramount as the full
implementation of the inflation targeting framework requires a deep
understanding on the work of the monetary policy in the economy both
in the short-run and long-run.
This paper is a part of the research agenda on the monetary
transmission mechanism in Indonesia. In the recent years, a large body
of literature has been developed on the efficacy of monetary policy and
channels through which the policy affects the real economy. Traditionally,
monetary policy is believed to influence the economy through money or
short-term interest rate which in turn affect the long term interest rate and
cost of capital and thus investment. For example, in a monetary
3
contraction, the banks reserves decrease and due to the reserve
requirement the ability of banks to issue the deposits were constrained. As
a result, the depositors hold less money (bank deposits) in their portfolios.
If prices are sticky, real money balance will fall and both short-term
interest rates and (through expectational effect) the *long-term interest
rates will rise. Accordingly, demand for loans, investments and interest-
sensitive spending such as housing, all fall.
For the last decade, there have been growing a voluminous studies
of the effects of imperfection in financial markets on the real economy
and business cycles (see e.g. Gertler, 1988; Bernanke and Gertler, 1989).
The understanding of the role of the financial market imperfection has
also generated theories on monetary transmission mechanism which
emphasize the importance of this imperfection, especially asymmetric
information problem in credit market, in explaining the effects of a
monetary policy. These theories can be categorized as the „asymmetric
information based transmission mechanism‟ or credit channel. There are
two strands of literature on the credit channel. First, the bank lending
channel which emphasize the effects of monetary policy on bank
balance sheet, especially in the asset side of banks. Second, the balance
sheet channel which emphasize the effects of monetary policy on firm
balance sheet and thereby access to banks‟ credit.
According to the bank lending channel, banks participate in the
transmission of monetary policy not only via their liabilities side but also
through their assets. For example, in a monetary contraction, banks
reserves decrease and owing to reserve requirements, bank deposits fall.
Should the decrease in bank deposits be not offset by other funds which
are free from reserve requirements, or by a decrease in securities, the
consequence would be a fall in bank loans. If bank loans also fall and
4
bank dependent borrowers are dominant in the economy, the restrictive
monetary policy results in a fall in investment and economic activity.
Hence, monetary policy not only directly influences the real interest rate
but also directly affects the supply of bank loans. Thus, two necessary
conditions for the existence of this channel are: (1) bank loans and
securities must be imperfect substitutes for some borrowers, or some
borrowers are bank dependent; (2) the central bank must be able to
constrain the supply of bank loans. While the first condition is likely to be
satisfied as the bank lending is still the dominant source of funds for firms‟
financing, the second condition is subject to empirical investigation. Using
sample data from 1985-1995, Agung (1998) proved that a monetary
policy was able to influence the bank supply of credit, in particular, of
small banks, not large banks which were able to shield their bank loan
supply by finding the cheaper source of funds from abroad.
This paper investigates the bank lending channel of monetary
transmission using the sample data including the period after the crisis and
using various tools to analyse. This is stimulating, at least for two grounds.
First, the evidence of ability of large banks to protect the lending supply
by accessing non deposit funds from abroad may be questioned recently
as the access to foreign funds has been very limited. Second, the
existence of credit crunch (Agung, et al., 2001) supports the bank lending
channel, i.e., the credit market is more supply-determined, rather than
demand-determined as suggested by the money/interest-rate channel.
However, the existence of credit crunch in which the non-price rationing
exist, simultaneously shows that the effectiveness of monetary in
influencing the supply of credit has also been reduced.
A battery of tests will be utilised to analyse the bank lending
channel. First, we use VAR approach as Bernanke and Blinder (1992) using
5
aggregate data to see effects of monetary policy on bank balance
sheets. However, the empirical studies using aggregate data suffer from
identification problem: the inability to establish whether the decline in
credit as a result of a monetary contraction stems from a decline in loan
supply or driven by the fall in demand for loans as a result of the high long
term interest rate as predicted by the interest rate channel. Accordingly,
following Kashyap and Stein (1995) we also use disaggregated data to
deal with this identification issue. The use of the disaggregated data,
hypothesis underlying the bank lending channel can be analysed. That is,
following a monetary contraction, smaller banks which do not have
access to other source of funds will decrease their loan supply more than
that of large banks. On the borrower side, small borrowers which
presumably are characterized by stronger informational asymmetries and
lower access to alternative source of funds should be more sensitive to
monetary contraction (Gertler and Gilchrist, 1993, 1994). As
complementary to the VAR analysis, we estimate long-run demand and
supply equation of the Indonesian credit market derived from vector error
correction model (VECM) following Kakes (2000) in order to identify
whether adjustment toward the equilibrium in the credit market is
dominated by supply as suggested by lending channel. Finally,
disaggregated evidence is analysed using bank level panel data to
examine whether a monetary shock generates differential effects across
banks according to their net worth (capital) position.
The remainder of this chapter is organised as follows. Section 2
reviews the role of banks in the monetary transmission mechanism. Section
3 presents the empirical results of the VAR analyses. Section 4 reports on
the empirical results of the credit market model for Indonesia. Section 5
6
presents the evidence from panel data regression. Section 6 reports the
survey results. Finally, a summary and conclusions is presented in section 7.
2. The role of banks in monetary transmission
There is widespread agreement among economists that banks or
financial intermediaries have generally played an important role in
transmitting monetary policy to the real economy. But the precise the
role of banks is still debated. In the standard view, known as the money
or interest rate channel, banks play a special role on the liabilities side, i.e.,
the banking system creates money (liquidity) by issuing deposits1 and plays
no role on the assets side. In a monetary contraction, the banks reserves
decrease and due to the reserve requirement the ability of banks to issue
the deposits were constrained. As a result, the depositors hold less money
(bank deposits) in their portfolios. If prices are sticky, real money balance
will fall and both short-term interest rates and (through expectational
effect) the long-term interest rates will rise. Accordingly, demand for loans,
investments and interest-sensitive spending such as housing, all fall.2 So,
three crucial conditions must be satisfied for the existence of a money
channel are: (1) prices are sticky so that monetary policy can affect real
money balances, (2) short-term interest rates do influence long-term
interest rates; and (3) the latter do influence real investment expenditure.
According to the “bank lending” (Bernanke and Blinder, 1988)
monetary transmission mechanism, banks‟ assets as well as their liabilities
1 By making loans or buying bonds. 2 Alternatively, some have described the money view as the standard IS-LM model, thus
it does not require the role of banks. In the IS-LM, if banks did not exist, central banks
could buy and sell bonds from public. This would influence interest rates and hence real
investment expenditure.
7
play an important role. In a monetary contraction, banks‟ reserves
decrease and given reserve requirements, banks‟ deposits fall. If the
decrease in deposits is not offset by other funds which are not subject to
reserve requirements, or by a decrease in securities, this will result in a
decrease in bank loans. If bank loans fall and bank dependent borrowers
are dominant in the economy, real investment expenditure will fall. Since
bank loans in many countries, especially developing countries, remain the
main source of external finance for business enterprises, a disrupting of
bank loan supply can reduce the economic activity. The necessary
conditions for the existence of this channel are: (1) the central bank must
be able to constrain the supply of bank loans, (2) bank loans and
securities must be imperfect substitute for some borrowers.
In regard to the second condition, since asymmetric information in
financial markets in developing countries seems to be prevalent, some
class of borrowers find it difficult to issue securities. Banks play an
important role in overcoming the information problem in credit markets,
consequently many borrowers are substantially bank dependent.3 Thus
the second condition seems to be satisfied. As pointed out by Bernanke
and Gertler (1995), the first condition is questionable in empirical grounds,
that is, whether monetary policy can significantly influence the supply of
bank loans. As we discussed above, in order to limit the ability of banks
to extend their loans after monetary contraction, banks must not easily
issue another form of liabilities to replace lost deposits. In other words, all
components of bank liabilities (except capital) must be subject to reserve
3 Even in the countries like US where financial markets have been well established, the
number of bank dependent borrowers are substantial (see Himmelberg and Morgan,
1995).
8
requirement4. Of course, to some extent banks which have gone public
can issue new equity to generate loanable funds.5 However as argued
by Kashyap and Stein (1995), as long as the banks do not face a
perfectly
elastic demand for their managed liabilities, a bank lending channel will
operate. Some argue that the regulatory action of central banks can
also significantly influence bank loan supply. For example, much research
has focused on the effect of capital adequacy regulations on the banks‟
willingness to lend (see for example, Peek and Rosengren,1995a,b).
The third channel through which monetary policy might be
transmitted to the real economy is known as the balance-sheet channel.6
The basic idea of this theory is that monetary policy can affect borrower‟s
financial position or collateralisable net worth, thereby influencing the
costs of external finance, which in its turn affects a borrower‟s financial
decision to engage investment expenditure. Suppose central banks
conducts a monetary tightening which raises market interest rates. This
can directly influence the borrowers‟ financial position in two ways.7 First,
it can cause a deterioration of the assets prices of the borrowers, and
hence reduce the value of the collateral the borrowers hold. Second, an
increase in market interest rates raises the cost of servicing outstanding
short-term or floating debt and so reduces net cash flow. Due to
asymmetric information, moral hazard problem and bankruptcy law,
borrowers with lower net worth are less creditworthy since the lender must
4 However, even if the reserve requirement (RR) is equal over all class of liabilities, in
practice, RR can not be flexibly applied. Hence, at least in the short run banks can
escape from monetary tightening. 5 Since 1989 banks in Indonesia have been allowed to issue equity. 6 Some call it the “broad credit channel” or “net worth channel”. For theoretical
exposition of this channel, see Bernanke and Gertler (1989), Calomiris and Hubbard
(1990), and Bernanke, Gertler and Gilchrist (1996). 7 See Bernanke and Gertler (1995) for detailed discussion.
9
bear higher costs in the event of the project fails. In contrast, the higher is
a borrower‟s net worth, the greater is his collateral, hence the lower are
the monitoring costs borne by the lenders. Consequently, the lenders
impose a varying premium for external finance reflecting the cost of
monitoring and evaluation. Thus, under information asymmetries, the
internal finance of a new project is cheaper than external finance; and
monetary policy influences the „wedge‟ between the internal and
external finance. A monetary contraction will increase the wedge and
vice versa.
The second and third channels have similarities. Both theories
suggest that monetary policy can influence borrowers who have a limited
access to capital markets, and the bank dependent and the finance-
constrained borrowers to some extent are identical. The differences
between the two channels is that in the bank lending channel the
condition that monetary policy must be able to affect bank loan supply is
crucial. In contrast, in the balance sheet channel banks are not the
central player, instead what matters is any disturbance which can
influence the premium paid for external finance. Since the arguments of
these two channels are primarily based on credit (capital) market
imperfection, they two channels are often classified as the „credit
channel‟ or the „capital market imperfections channel‟.
3. Empirical Evidence from VAR Analysis
3.1. Data
We analyse monthly data over a sample that runs from 1991:01 to
2000:12. Most earlier studies of the bank lending channel employed
aggregate data, comparing the relationship between total bank loans
10
versus total deposits and the economic variables in the context of vector
autoregressions (see Bernanke and Blinder, 1992) or the relative
forecasting power of the two aggregates with respect to output
fluctuations (Ramey, 1993, Kim, 1999, among others). However, it is now
widely agreed that testing with aggregate data can generate a
misleading conclusion. First, the use of aggregate time series cannot
resolve the well-known identification problem, i.e. to distinguish whether
the credit contraction which typically follows the monetary tightening is a
result of the supply by banks, as argued by the bank lending channel, or
the fall in demand for bank loans stemming from a recession. Second,
testing the relative importance of the bank lending vs the money view by
comparing the information content of these two aggregates with respect
to output would be misleading (Bernanke, 1993). Due to bank balance
sheet constraints, aggregate money supply (liability side of banks) and
aggregate bank loans (assets side of banks) by construction, move
together although they are not identical. Thus the relative forecasting
power of these two aggregate variables does not provide any
information about monetary transmissions.
To identify the channel of monetary policy, recent studies (Kashyap
and Stein, 1995, 1997; Dale and Haldane, 1995, Kakes, 2000, for example)
have tended to use cross sectional data to determine whether there are
distributional effects of monetary policy across lenders and borrowers, as
predicted by the bank lending channel argument. On the lenders side,
the lending view suggests that a monetary policy shock should constrain
bank loan supply since banks cannot frictionlessly raise non-deposit funds
to make up for a shortfall in their deposits. But this will depend on the
ability of banks to insulate themselves from the shock. Small banks which
have relatively limited access to non-deposit funds such as securities issues
11
or foreign borrowings are expected to be more affected by the monetary
shock and to tend to cut their loan supplies immediately following the
shock. On the borrower side, small firms that have limited access to
external finance should be more sensitive to a monetary shock (Gertler
and Gilchrist, 1994). The use of cross sectional data, furthermore,
eliminates the banks‟ balance sheet constraints.
This study follows Kashyap and Stein (1995) by disaggregating banks
into different classes, reflecting their size and accessibility to non-deposit
funds: state banks which are large and foreign exchange licensed banks,
private banks and foreign and joint venture banks. The source of data is
from Banking Statistics Monthly Report. The data for each class of banks
include bank loans, deposits, non-deposit funds and securities holdings.
The loans are also disaggregated into class of borrowers, i.e. loans to
individual and private enterprises and disaggregated into different types
of use, i.e., investment and working capital (Appedix A provides detailed
definition of data used).
3.2. VAR specification
The effects of monetary policy shock on bank balance sheets and
economic variables are examined using the vector autoregression (VAR)
approach. Specifically, we use the standard semi-structural VAR
approach as suggested by Bernanke and Blinder (1992) instead of
structural VAR since we do not explicitly use a theoretical framework to
identify the innovations, but impose a causal ordering. A structural model
is a linear dynamic system of the following form:
By C L yt t t ( ) (1)
12
or in MA form:
y Lt t ( ) (2)
where (L)=[B-C(L)]-1. y is nx1 vector of endogenous variables in the
system including one policy variable and some non-policy variables. t is a
vector of structural shocks, including the monetary policy shock. B
represents the structural parameters of contemporaneous endogenous
variables and C(L) is kth degree matrix polynomial in the lag operator, i.e.
C(L) = C1L+C2L2+...+ CkL
k. t is an nx1 vector of structural shocks with zero
mean, orthogonal and variance-covariance matrix E(tt) = I.
Equation (1) can be written in a reduced form which can be
estimated by OLS as:
y A L y ut t t ( ) (3)
with E(utut) = . By noting A(0) = B-1 , from the structural model (1) and the
reduced form model (3) we obtain:
A(L)=A(0)C(L) (4)
and,
ut = A(0)t (5)
Accordingly,
E(utut) = = A(0)A(0) T (6)
From (2) we can obtain impulse-response functions, (L), to structural
shocks, t and (L) can be calculated from (3) and (5):
(L) = [I-A(L)]-1
A(0) (7)
In order to identify the structural model and structural shock, t we have to
determine the nxn elements of matrix A(0). As is known from OLS
13
estimates of (3) we can solve (6) for A(0) and then deduce t from (5).
However system (6) only provides n(n+1)/2, hence we need n(n-1)/2
additional restrictions for the identification. A convenient way to add the
n(n-1)/2 restrictions is to assuming A(0) is lower triangular and use the
Cholesky decomposition of the variance-covariance matrix (Sims,
1980). This restriction is equivalent to assuming that the residuals ut form a
recursive system. The ordering of the variables in the system, therefore,
affects the recursive chain of causality among the shocks in any given
period. The policy variable is placed first (for example, Sims, 1992) if we
assume that there is no contemporaneous feedback from non-policy
variables onto the policy variables. Thus, this equivalently assumes that
the monetary decisions are made without considering the simultaneous
evolution of economic variables. This assumption is plausible if data of
non-policy variables are not readily available. If we assume that the
policy variable responds to contemporaneous feedback from non-policy
variables but there is one period lag of feedback of the policy shock on
non-policy variables, the policy variable should be placed last. Given the
high frequency data (monthly) that we use in constructing VARs, hence
the existence of information lag from non-policy variables8, we prefer the
former identifying restriction. Nevertheless, as the correlations across
residuals (t) are very small, the ordering is actually not significant.9
In contrast to Agung (1998), in examining the effects of monetary
policy on the bank balance sheet, we include all bank balance sheet
components in a VAR, so that the interrelationship between the balance
sheet components can be evaluated. This approach follows McMillin
8 For example, interest rate data (the policy variable) is readily available, while non-
policy variables such as real output and price were available with a lag. 9 A rule of thumb is that if ij< 0.2 for ij, the ordering of variables in a VAR is not
relevant (Enders, 1995, pp.309).
14
(1996). Since such specification involves a VAR with many variables, while
our data is rather limited, we use only lag of 3. The systems we developed
are six-variable VARs with the following ordering: monetary policy
indicator, bank deposits, loans, output and prices. We use real GDP and
deflator GDP for output and prices, respectively, when aggregate loans
are used. In the VAR using disaggregated data, for loans to individual, we
use real consumption and consumer price index as the output and prices,
while for loans to private enterprises, we use the production index and the
wholesale price index. In the disaggregation based on the bank
categories, the balance sheet variables are deposits and loans of the
relevant banks.
Table 1. Unit Root Tests
Variabel Level First Diff.
Lag ADF Test Lag ADF Test
SBI Rates 6 -2.851*** 11 -3.771*
Interbank Rates 1 -2.196 1 -10.361***
Base Money 11 -1.549 11 3.347**
Exchange rate 9 -1.982 8 -3.783***
Real GDP 10 -2.018 12 -2.035
GDP Deflator 9 -1.896 6 -3.622***
Consumer PI 8 -1.603 7 -4.022***
Real Consumption 12 -0.841 11 -3.427**
Deflator Consumption 8 -1.489 7 -3.847***
Production Index 9 -1.319 8 -3.103**
Real Investment 7 -1.352 6 -7.295***
Deflator Investment 11 -1.746 10 -2.831***
Deposit - Comercial Banks 12 -2.501 1 -7.283*
Deposit - State Banks 2 -1.300 1 -6.285*
Deposit - Private Banks 1 -1.182 10 -3.529**
Deposit - FX Banks 11 0.361 10 -3.147***
Deposit - Foreign Banks 1 -3.067 1 -10.122*
Lending - Comercial Banks a 1 -1.269 12 -2.532
Lending - State Banks a 8 -1.011 9 -2.095
Lending - Private Banks 2 -0.888 9 -3.256**
Lending - FX Banks a 8 -3.062 1 -5.840*
Lending - Foreign Banks 10 -0.488 9 -2.825
15
Total Lending To Private Enterprises 8 -3.193* 8 -2.090
Lending To Individuals 3 -1.671 2 -3.528***
Inv.Lending Rates - Commercial.Banks 1 -1.646 1 -10.516***
Inv.Lending Rates - State Banks 7 -2.352 1 -9.583***
Inv. Lending Rates - Private Banks 8 -3.091** 11 -2.933**
Inv.Lending Rates - Foreign Banks 9 -3.162** 7 -3.288**
Invest. Lending - Commerc.Banks 8 -3.594** 12 -3.231**
Invest. Lending - State Banks 8 -3.266* 9 -2.951**
Invest. Lending - Private Banks 8 -2.701 8 -2.497
Invest. Lending - Foreign Banks 2 -3.163* 4 -4.325***
WC.Lending Rates - Commercial.Banks 10 -2.839* 9 -2.977**
WC.Lending Rates - state banks 7 -2.738* 9 -3.228**
WC.Lending Rates - Private Banks 7 -2.826* 9 -2.484*
WC.Lending Rates - Foreign Banks 4 -2.488 9 -3.211**
Work. Cap. Lending - Commerc.Banks 8 -2.791 4 -3.036**
Work. Cap. Lending - State Banks 5 -1.711 1 -8.660***
Work. Cap. Lending - Private Banks 1 -1.577 1 -6.253***
Work. Cap. Lending - Foreign Banks 8 -3.685** 12 -3.941***
Notes : For levels, time trend and constant were included in the tests, while for first-difference, only constant was
included.
Critical values: levels: 5 % (*) = -3.44, 1%(**)=-4.02; first-differences: 5%(*)=-2.88, 1%(**)=-3.47. a First difference of
these variables are significant at 10% without trend and constant
All variables are in log levels except for interest rates and were
tested for stationarity by Augmented Dickey Fuller (ADF) tests (see Table
1). In general, the results indicate that all were found to be I(1). In spite of
non-stationarity of data, Sims (1980) and Doan (1992) do not recommend
differencing the data prior to VAR estimation even if they contain unit
roots. Their argument is that differencing in order to assure stationarity will
„throw away‟ valuable information concerning the interrelationships of
the variables in the system such as the possibility of cointegrating
relationship. It should be noted that the emphasis of VAR analysis is to
trace the dynamic relationships among a set of interested variables, not
the parameter estimates. Therefore, the VARs were estimated with all
variables in levels. Tests of cointegration using Johansen maximum
likelihood suggest that all systems are cointegrated (Appendix B).
16
3.3. Measuring the monetary policy variable
The crucial part of the studies on transmission mechanism is how to
measure the monetary policy indicator. The literature on the identification
of monetary policy indicator suggests that there are some alternatives to
measure the indicators: the interest rate used by the central bank to
influence the money market such as the Federal Funds rate in US
(Bernanke and Blinder, 1992), Romers‟s dates of monetary tightness
(Romer and Romer, 1990), or some aggregates such as base money, total
reserves, non-borrowed reserves (Strongin, 1995 and Christiano,
Eichenbaum and Evans, 1996).
Agung (1998) uses the money market interest rate (interbank money
market) as the monetary policy variable by arguing that Bank Indonesia
often indirectly targets the interbank interest rates. An alternative is the
SBI rates which have been widely used as the benchmark by the market,
in particular since the banks‟ holding of SBIs increased dramatically. The
problem of using the SBI rates are the auction system has been changed
three times. Before 1993, Bank Indonesia targeted the quantity of SBI in
the auction (cut-off rate), but since 1993 the system was changed to stop-
off rate in which the monetary authority set the interest rates on SBIs and
market determines the quantity of SBIs. The stop-off rate system was
changed again into cut-off rate in 1998. In practice, however, a mix of
price and quantity targets has been frequently executed. Another
alternative is base money, which has formally been used by Bank
Indonesia as the operating target since 1998.
In order to choose the appropriate measure of monetary policy we
follow a simple approach suggested by Bernanke and Blinder (1992). In
this approach, the selection is based on the information content of the
policy variables in the reduced forms of various real variables such as real
17
GDP, consumer price index, production index, consumption and
investment. In addition to lags of the real variables and the policy
variables, the reduced form also incorporates other monetary aggregates
such as M1. We use lag of 12 for all independent variables in the reduced
form. Since the variables are I(1), an alternative specification is the
models are specified in the ECM model in which the variables are
specified in first differenced and lag of cointegrating relationship is
included in the model. Using this approach we can select the policy
variables based on the short run information content (significance of the
first differenced) and the long run information content (significance of the
ECM coefficient). The results suggest that in general SBI rates and PUAB
rates perform better than base money. However, it is difficult to select the
best policy indicator between the two interest rates. With regard to the
short run information content, PUAB rates perform better than SBI rates,
while in the ECM specification, the long-run information content of SBI is
superior to the PUAB rates. Hence we use the two policy indicators and
compare the results in the VARs.
Table 2. Information content of the monetary policy variables
Variables Level Level Level
Spec. First Diff. ECM Spec. First Diff. ECM Spec. First Diff. ECM
GDP 0.151 0.389 -0.010 0.016 0.009 -0.104 0.042 0,025 -0.034
(2.027) (3.394) (3.964)
Indeks Produksi 0.319 0.248 -0.053 0.298 0.229 -0.005 0.000 0.000 0.002
(2.143) (1.267) (1.569)
CPI 0.742 0.564 0.001 0.004 0.004 -0.053 0.030 0.009 0.018
(0.032) (1.351) (0.724)
ECM Spec. ECM Spec. ECM Spec.
Base Money SBI Rates Interbank Rates
3.4. Innovation Analysis
Figure 1 and 2 reports the impulse responses of variables in the VAR
to a monetary shock measured by the SBI rates for the whole period and
18
period before the crisis. Generally speaking, the adverse effects of a
monetary tigntening on the banks‟ balance sheet and macroeconomic
variables are much stronger than those before crisis.
Figure 1. Effects of a Monetary Shock (SBI rate)
A. Whole sample B. Before crisis
-.03
-.02
-.01
.00
.01
.02
.03
5 10 15 20 25 30
SBI RATE
-.02
-.01
.00
.01
5 10 15 20 25 30
DEPOSIT
-.05
-.04
-.03
-.02
-.01
.00
.01
.02
.03
5 10 15 20 25 30
LENDING
-.08
-.04
.00
.04
.08
5 10 15 20 25 30
SECURITIES
-.020
-.015
-.010
-.005
.000
.005
.010
5 10 15 20 25 30
GDP REAL
-.03
-.02
-.01
.00
.01
.02
.03
5 10 15 20 25 30
GDP DEFLATOR
-.002
-.001
.000
.001
.002
.003
.004
.005
.006
5 10 15 20 25 30
SBI RATE
-.008
-.004
.000
.004
5 10 15 20 25 30
DEPOSIT
-.004
-.002
.000
.002
.004
5 10 15 20 25 30
LENDING
-.06
-.04
-.02
.00
.02
5 10 15 20 25 30
SECURITIES
-.005
-.004
-.003
-.002
-.001
.000
.001
.002
.003
5 10 15 20 25 30
GDP REAL
-.003
-.002
-.001
.000
.001
.002
.003
.004
.005
5 10 15 20 25 30
GDP DEFLATOR
Before the crisis, bank lending is almost not affected by a tight
monetary policy. This result is consistent with findings by Agung (1998)
who also use pre-crisis data. One of a reasonable explanation of low
sensitivity of lending to a monetary shock is that before the crisis,
especially since the beginning of 1990s, the access of domestic
commercial banks to international source of funds was relatively easy.
Hence, in spite of tight money, they can still provide loans to their
borrowers. A survey conducted by Hadad (1996) also found similar
phenomenon. During the tight money period (e.g. in the aftermath of
what so called-Gebrakan Sumarlin), the loan growth of state banks and
19
large private banks was higher than their deposit growth. In fact,
domestic banks have been major issuers of bonds into international
markets during the period (World Bank, 1996). Large banks obviously have
better credit ratings than smaller banks and are thus able to raise funds
less expensively. This differential behaviour of state banks and private
banks is clearly reflected in Figure 3. Loans of state banks are completely
insensitive to a monetary shock, while that of private banks are more
sensitive to the shock.
The relatively high sensitivity of commercial bank lending for the
whole sample is partly influenced by behaviour of bank lending during
and after the crisis. Given weakening of firms‟ balance sheet amidst low
economic prospect, a monetary tightening worsens the firms‟ financial
position and raises the probability of default and hence reduces the
willingness of bank to lend. This is consistent with a recent study by
(Agung et al., 2001) who found the existence of „credit crunch‟ in the
aftermath of the crisis. Under such situation, they argue, a tight money
exacerbates the unwillingness of banks to lend. This is also confirmed by a
corresponding study on balance sheet channel that concludes the
existence of financial accelerator effect of monetary policy, especially
after the crisis. Similar impulse responses is obtained if we use the PUAB
rate as the policy variable, although the effect of a change in SBI rate
seems to be more pronounced than a change on PUAB rate.
For the whole period sample, although the bank lending is
responsive to a monetary shock, its response is rather slow, i.e. about 10
months for bank lending to decline after a shock occurs. Another bank
asset portfolio, the securities holdings of commercial banks, immediately
fall after a shock and take about 10 months to return back. This
behaviour can be interpreted as an indication that banks prefer to use
20
their securities holdings as a buffer stock to offset monetary shocks. This
behaviour is consistent across different types of banks, except for the
foreign and joint-venture banks (Figure 3). This is not surprising that after a
monetary contraction there is an indication of flight to quality of deposits,
especially from private domestic banks to foreign banks. While deposits
of private banks fall immediately after a shock, deposits of foreign banks
increases (Figure 3). Accordingly, the foreign banks have an ample
deposit funds to maintain the credit line without liquidating their securities
holdings.
Figure 2. Effects of a Monetary Shock (PUAB rate)
A. Whole sample B. Before crisis
-.04
-.02
.00
.02
.04
.06
.08
5 10 15 20 25 30
INTERBANK RATE
-.02
-.01
.00
.01
.02
.03
.04
5 10 15 20 25 30
DEPOSIT
-.08
-.04
.00
.04
5 10 15 20 25 30
LENDING
-.10
-.05
.00
.05
.10
.15
.20
5 10 15 20 25 30
SECURITIES
-.03
-.02
-.01
.00
.01
5 10 15 20 25 30
GDP REAL
-.02
-.01
.00
.01
.02
.03
.04
.05
5 10 15 20 25 30
GDP DEFLATOR
-.002
.000
.002
.004
.006
.008
.010
.012
5 10 15 20 25 30
INTERBANK RATE
-.008
-.004
.000
.004
.008
5 10 15 20 25 30
DEPOSIT
-.002
-.001
.000
.001
.002
.003
.004
.005
.006
5 10 15 20 25 30
LENDING
-.06
-.05
-.04
-.03
-.02
-.01
.00
.01
.02
.03
5 10 15 20 25 30
SECURITIES
-.006
-.004
-.002
.000
.002
.004
.006
5 10 15 20 25 30
GDP REAL
-.002
-.001
.000
.001
.002
.003
.004
.005
.006
5 10 15 20 25 30
GDP DEFLATOR
The lag of bank lending to respond a shock can be attributed to
the fact that bank lending practices, especially investment loans, are
mostly conducted on a loan commitment basis, instead of on a project or
21
fixed-term basis. Under such a commitment, banks allow borrowers to
draw down a line of credit at their discretion; and borrowers pay a fee for
the credit line and pay interest on actual loans that have been drawn. As
a result of this system, banks cannot prevent the borrowers from drawing
the credit even when the monetary condition is tightened. Banks can
only reduce the supply of new loans, which presumably does not
immediately lead to a substantial fall in aggregate loans.
Figure 3. Effects of a monetary shock to balance sheet of different types
of banks: the whole sample
A disaggregation of total bank loans into corporate lending and
individual (household) lending (Figure 4), however, suggests that the
insignificant response of aggregate lending stems from the loan to firms.
By contrast, the loans for individuals drop significantly in the aftermath of
monetary shock. This may be explained by what so-called the „flight to
quality‟ phenomenon as suggested by Bernanke, et al. (1996). That is, in a
monetary contraction, to compensate the decline in cash flow, the
-.03
-.02
-.01
.00
.01
.02
.03
.04
.05
5 10 15 20 25 30
DEPOSIT STATE BANKS
-.08
-.04
.00
.04
.08
5 10 15 20 25 30
LENDING STATE BANKS
-.2
-.1
.0
.1
.2
.3
.4
5 10 15 20 25 30
SECURITIES STATE BANKS
-.03
-.02
-.01
.00
.01
.02
.03
.04
.05
5 10 15 20 25 30
DEPOSIT PRIVATE BANKS
-.08
-.04
.00
.04
.08
5 10 15 20 25 30
LENDING PRIVATE BANKS
-.2
-.1
.0
.1
.2
.3
.4
5 10 15 20 25 30
SECURITIES PRIVATE BANKS
-.03
-.02
-.01
.00
.01
.02
.03
.04
.05
5 10 15 20 25 30 35
DEPOSIT FX BANKS
-.08
-.04
.00
.04
.08
5 10 15 20 25 30 35
LENDING FX BANKS
-.2
-.1
.0
.1
.2
.3
.4
5 10 15 20 25 30 35
SECURITIES FX BANKS
-.03
-.02
-.01
.00
.01
.02
.03
.04
.05
5 10 15 20 25 30
DEPOSIT FOREIGN BANKS
-.08
-.04
.00
.04
.08
5 10 15 20 25 30
LENDING FOREIGN BANKS
-.2
-.1
.0
.1
.2
.3
.4
5 10 15 20 25 30
SECURITIES FOREIGN BANKS
22
creditworthy borrowers have access to short-term loans, while loans to the
less creditworthy borrowers such as individuals or small firms will be
rationed.
Figure 4. Effects of a Monetary Shock (SBI rate): Corporate vs Individual
Loans
A. Lending to Corporate B. Lending to Individuals
-.06
-.04
-.02
.00
.02
.04
.06
5 10 15 20 25 30
LENDING TO FIRMS
-.06
-.04
-.02
.00
.02
.04
.06
5 10 15 20 25 30
LENDING TO INDIVIDUALS
Figure 5. Effects of a Monetary Shock (SBI rate): Working capital vs
Investment Loans
A. Investment Loans B. Working capital Loans
-.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 .05
5 10 15 20 25 30
-.04 -.03
-.02 -.01 .00
.01
.02
.03
.04
.05
5 10 15 20 25 30
23
4. Evidence from Adjustment in the Credit market
4.1. Methodology
The hypothesis of the bank lending channel is that a monetary
policy affects the supply of bank lending which in turn influence the
investment by bank-dependent borrowers. The crucial assumption
underlying the hypothesis, thus, the credit market is supply-determined.
Following the approach conducted by Kakes (2000), we utilize the
Johansen‟s cointegration framework and impose the restrictions on the
cointegrating parameters representing the long run supply of and
demand for credit and to examine whether the short run adjustment
toward equilibrium is dominated by the demand or supply determined. If
the system is dominated by supply, one would expect that initially the
adjustment mainly takes place in the direction of the supply equation,
while eventually both relationships are satisfied.
The model of credit market is developed in a four-variable VECM.
The supply of bank lending is a function of the spread between the bank‟
lending rate and bank‟ funding costs proxied by the deposits rate, and
the level of economic activity. Whereas, the demand for loans is a
function of the lending rate and the level of economic activity. Thus, the
VECM includes the following variables: bank lending, level of economic
activity, loan rate and deposit rate. We analyse a VECM of the working
capital credit and investment credit markets as well as the disaggregation
according to bank categories.
In testing the Johansen‟s cointegration framework, we use the so-
called Pantula principle (Johansen, 1991) to select the deterministic
components and the rank of the cointegration matrix . The results
24
suggest that the rank of r = 2 which implies that we have to find two long-
run relationship in order to identify the cointegration space. Table 3 also
shows that we use model 1, i.e., include a constant in the cointegration
relationship and allow for a trend in the levels of variables.
Table 3. Selection of the cointegration model: joint test of deterministic
specification and rank of long run matrix
Variables
trace
Model 1 Model 2 Model 3
Working Capital
r = 0 51,37** 49,52** 55,82**
r 1 33,33** 32,55** 34,80**
r 2 9,567 9,083 23,37*
r 3 2,392 0,495 9,074
Investment
r = 0 40,51** 38,55** 44,39**
r 1 31,05** 24,93* 26,39*
r 2 12,360 11,67 11,74
r 3 6,955 5,94* 7,419
4.2. Results
Table 4 provides the unrestricted cointegrating relationship, given
rank r = 2. The first cointegrating equation likely represents the supply
function, while the second cointegrating equation represent the demand
function. Table 4 also report the outcome of restriction. We impose two
restrictions: equality restriction in the supply equation, i.e., coefficients on il
equals to (-) coefficients on id and exclusion restriction, coefficient on id is
zero in the demand equation. After normalizing with respect to Lt, the
following long-run supply and demand relationship, respectively:
25
Market of working capital credit:
)(046.0166.2 d
t
l
tt
S
t iiyL (11)
l
tt
D
t iyL 016.0792.1 (12)
Market of investment credit:
)(008.0567.2 d
t
l
tt
S
t iiyL (13)
l
tt
D
t iyL 011.0473.2 (14)
The credit demand equation shows long-run income elasticities of 1.8 and
2.5 for working capital and investment loans, respectively, which are
comparable to studies in other countries (e.g. Kakes, 2000, Fase, 1995).
The interest elasticities are -0.36 (-0.016 x 22.5, i.e. product of semi-elasticity
of loan rate and sample mean of loan rate) and -0.20 (-0.011 x 18.18) for
working capital and investment loans, respectively. These results are very
intuitive that the demand for working capital loans (short term loans) is
more sensitive to a change in the loan interest rates. Similarly, the
elasticity of interest rate for working capital loans in the supply function is
higher than that for investment loans.
Table 4 shows the adjustment coefficients, , which indicate the
speed towards the long run equilibrium. Compare coefficient for supply
and demand equation for working capital loans, it appears that bank
loans adjust significantly in the direction of the long-run supply of credit (
= - 0.10), while the adjustment to the demand equation is insignificant with
speed of adjustment coefficient 0.04. The same conclusion is also found
for investment loans, i.e., coefficient of speed of adjustment for supply
equation is – 0.28, while that for demand equation is 0.19. This suggest
that in the short run the market for working capital and investment credit is
dominated by supply rather than demand.
26
Table 4. Results for cointegration and restricted cointegration
A. Working Capital Loans
[1] Test for the number of cointegrating vector
L Y i L i D
Eigenvalues 0,424407 0,301216 0,0977544 0,025388
Hypothesis r = 0 r 1 r 2 R 3
trace 96,66** 45,29** 11,96 2,39
max 51,37** 33,33** 9,57 2,39
[2] Standardized eigenvectors, ‟
L Y i L i D
Vector1 1,000 -6,079 -0,698 0,528
Vector2 -1,285 1,000 -0,238 0,161
[3] Standardized adjustment coefficient,
L Y i L i D
Vector 1 -0,016 0,002 0,134 0,004
Vector2 0,026 0,004 -0,001 -1,084
[4] Restricted eigenvectors
L Y i L i D
Supply (=-0.10) 1 -2.17 -0.05 0.05
Demand (=0.04) 1 -1.79 0.02 0.00
B. Investment Loans
[1] Test for the number of cointegrating vector
L Y i L i D
Eigenvalues 0,254408 0,201477 0,0857003 0,0491518
Hypothesis r = 0 r 1 r 2 R 3
trace 90,88** 50,37** 19,32 6,955
max 40,51** 31,05** 12,36 6,955
[2] Standardized eigenvectors, ‟
L Y i L i D
Vector1 1,000 -2,677 -0,022 0,013
Vector2 0,116 1,000 0,161 -0,066
[3] Standardized adjustment coefficient,
L Y i L i D
Vector1 -0,100 -0,009 0,313 0,172
Vector2 0,013 -0,011 -0,472 1,472
[4] Restricted eigenvector
L Y i L i D
Supply (=-0.28) 1 -2.57 -0.01 -0.01
Demand (=0.19) 1 -2.41 -0.01 0.00
27
5. Evidence from Bank Level Panel Data
As aforementioned previously, the existence of bank lending
channel should be tested using disaggregated bank data. In section 3
using VAR approach we have shown differential lending behaviour across
banks according to their accessability to non deposits funds and their
securities holdings to maintain their lending activities. While many studies
have used such VAR approach to investigate differential effects,
empirical evidence about the existence of bank lending channels using
bank-level data is scarce. Kashyap and Stein (2000) and Kishan and
Opiela (2000) for US and Bondt (1999) for European countries are the
notable exceptions. Using bank level data, Kashyap and Stein (2000)
found that sensitivity of bank lending to a monetary policy is determined
by buffer stock owned by banks. Lending of banks that have lower ratios
of cash and securities to assets are more sensitive to a monetary
tightening. Bondt (1999) using similar approach using bank level with
different size and liquidity to examine impact of monetary policy through
bank lending channel. Kishan and Opiela (2000) separate banks
according to their capital leverage ratio by arguing that capital‟s role in
absorbing shock to assets makes it an indicator of bank health and a
good indicator of bank‟s ability to raise funds during tight money policy.
In the following empirical approach, we combine both capital
(capital to asset ratio) and size (assets). We test empirically whether the
effects of monetary policy on bank lending are more pronounced for
small and low capital banks.
28
Methodology and Data
The empirical framework is as follows:
)*()*()*( 654321 ittitititititttiit AYYLCAPDEPDEPLCAPrrL
with index i referring to bank i and t to period t; L denotes log loans, r
denotes SBI or interbank (PUAB) rates, LCAP denotes low capital indicator
(1 for banks with Capital to Asset ratio below the sample first quartile and
0 otherwise), DEP denotes log deposits, Y denotes log real GDP, and A
denotes log total assets as a proxy of bank size.
We can expect that 1 < 0, that is, the an increase in interest rate on
SBI or PUAB, as the monetary policy indicator, will lead to a fall in bank
lending. The impact may differ across banks according to their capital
strength. The response of lending of low capital banks is expected to be
more sensitive to a change in interest rate, 2 < 0. Similarly, using deposits
as the proxy of the bank lending capacity (loanable funds), we can
expect that a higher the loanable funds the higher loan growth, 3 > 0.
The sensitivity of loan with respect to the loanable funds should be more
higher for low capital banks that may more difficult to find other sources
of funds, 4 > 0. Loan demand effects are assumed to be captured by
the growth rate of real GDP; higher economic activity will leas to a rise in
bank lending, 5 > 0. Assuming that bank and borrower size are positively
correlated, we can expect that the impact of loan demand to a
monetary shock may be stronger for small banks (Bondt, 1999) and hence
6 < 0.
Bank-level data are obtained from Monthly Bank Report, over
period 1994-1999. The sample size is 140 banks, all still exist until the end of
1999. In estimation, we split the sample estimation into before crisis, after
29
crisis and the whole sample to capture possible different behaviour. Table
5 summarizes the characteristics of data.
Table 5. Sample characteristics of banks data
Asset Capital Loan Deposit C/A
Before Mean 7,818,206 177,667 1,388,371 1,779,192 0.123
Crisis Median 446,002 41,013 220,792 140,900 0.101
1st quartile 149,001 15,815 81,088 51,403 0.063
After Mean 15,820,693 853,297 2,567,427 5,590,698 0.089
Crisis Median 1,131,292 61,182 420,213 301,733 0.071
1st quartile 299,345 22,947 82,115 94,578 0.033
Whole Mean 10,485,702 402,877 1,781,390 3,049,694 0.112
Period Median 575,147 51,009 248,567 187,872 0.092
1st quartile 180,021 17,410 81,421 61,322 0.051
Empirical Results
The results of estimation are presented in Table 6 and 7, for SBI and
PUAB rates as policy variables, respectively. Generally speaking, the
results are in line with our prior expectation. For the whole period, all
coefficients are significant. To test our hypothesis regarding the existence
of a bank lending channel, the significance level of a negative estimated
2, positive 4 is examined. The results support the hypothesis, that is, an
increase in interest of SBI or PUAB reduces the bank lending supply and
their effects are more pronounced for low capital banks. Furthermore, it
can be seen that bank lending supply is sensitive to lending capacity
available. What is more interesting result is that the sensitivity of bank
lending is higher for banks wit low capital. A significant negative
estimated 6 support the hypothesis that lending of small banks is more
sensitive to demand effect.
The next interesting findings are that there is differential behaviour
of bank lending before and after the crisis. Before the crisis, the interest
rate of SBI and PUAB do not significantly influence the bank lending. This
confirm to our VAR analysis above. However, during that period, lending
of banks with low capital is negatively affected by the monetary
30
tightening. Again, this result suggests that the bank lending channel is
operative through banks with low capital. This can also be interpreted as
a fact that bank lending channel is strongly working when the capital of
commercial banks is weak. After the crisis, the bank lending is sensitive to
a monetary shock though only significant at 10%. The sensitivities are also
higher for banks with low capital in spite of insignificant.
Table 6. Panel data results: SBI rate as the policy variable
Whole Sample
Before Crisis
After Crisis
Constant
0.01
(11.97)
0.01
(5.12)
-0.01
(-2.80)
r-sbit-1 -0.04
(-11.84)
0.002
(0.20)
-0.01
(-1.65)
r-sbit-1 * lowcap -0.01
(-3.14)
-0.03
(-4.18)
-0.01
(-1.20) dep 0.12
(37.19)
0.07
(18.80)
0.19
(30.62) dep * lowcap 0.09
(12.12)
0.03
(3.096)
0.07
(5.67) yt 1.24
(5.87)
1.15
(4.31)
0.89
(2.51) yt * Size -0.08
(-5.09)
-0.08
(-4.01)
-0.05
(-2.06)
Notes: Values in parentheses are t-statistics
Table 7. Panel data results: PUAB rate as the policy variable
Whole Sample (1994.01 – 1999.12)
Before Crisis (1994.01 – 1997.12)
After Crisis (1998.01 – 1999.12)
Constant
0.01
(10.14)
0.01
(8.73)
-0.01
(-3.29)
r-puabt-1
-0.03
(-9.89)
0.01
(1.57)
-0.01
(-1.07)
r-puabt-1 *
lowcap
-0.01
(-2.81)
-0.01
(-2.50)
-0.01
(-1.18) dep 0.13
(37.40)
0.07
(18.95)
0.195
(30.68) dep * lowcap 0.09
(12.09)
0.03
(2.75)
0.07
(5.68) yt 1.27
(5.95)
1.30
(4.89)
0.89
(2.48) yt * Size -0.08
(-5.54)
-0.09
(-4.54)
-0.05
(-2.14)
Notes: Values in parentheses are t-statistics
31
6. Evidence from survey
This section presents an analysis based on a survey on banks and
firms. The survey is designed to generate answers to some important
questions on behaviour banks and firms in the aftermath of a monetary
shock. From the banking survey, the main issue examined is whether
banks reduce their lending supply after a monetary shock, as expected
by the bank lending channel hypothesis. How do they reduce the
lending supply, by price or non-price mechanism? If they reduce their
loan supply with a lag, how do they maintain their lending supply? From
the firm survey, the issues examined are: what are sources of funds, how is
the sensitivity of demand for bank lending after a monetary tightening?
Are they rationed during a tight money periods?
6.1. Sample characteristics
The characteristics of the banks participating in the survey are
summarized in Figure 6. The number of banks interviewed was 28, which
can be categorized as follows: State Bank 14%, Private FX Bank 48%,
Private Non-FX Bank 7%, foreign & joint Venture Bank 32% and Regional
Bank 4%. According to asset size, the most respondents are relatively
large banks with around 65% of banks having more than Rp 10 triliuns.
Examining whether a bank is recapitalized after the crisis is important for
analysing their behaviour. Accodingly, we split the sample into
recapitalised banks (57%) and non-recapitalised bank (43%).
Private Non-
FX Bank
7%
Regional Bank
4%
State Bank
14%
Foreign &
Joint Venture
Bank
32%
Private FX
bank
43%
Rp10 triliun - Rp50
triliun
43%
More than Rp 50
triliun
18%
Less than Rp 1
triliun
14%
Rp1 triliun - Rp10
triliun
25%
Figure 6. Characteristics of Bank Sample
32
For the firm survey, we interviewed 141 companies, categorized
according to business sector and scale. According to business sectors,
the sample can be ctagorized into manufaturing sector (39%), trade
(31%), property and construction (14%), agriculture (6%) and other sector
(10%) (Figure 7). Meanwhile classified according to the size of turnover,
both large and medium firms have same portion 42%, but for small firm just
16% of total respondent. The majority of firms (63%) sell their product in
domestic market, and only 37% of firms have export orientation.
2. Are firms bank-dependent?
As outlined previously, the existence of the bank lending channel of
monetary transmission depends on whether the bank lending is a
dominant source of external funds. The survey indicates that in
conducting their business activities, the firms use internal fund as the main
source of financing (60,71%) (Figure 8). Meanwhile, as the source of
external financing, bank credit still serves as the main source of funds.
About 20,71% of firms use bank credits as the main sources of funds. This
finding is in line with a „credit crunch‟ survey conducted by Agung, et al.
(2001) and substantially different from results from surveys before the crisis
(e.g. Ang, Fatemi and Tourani-Rad, 1997). As found in many studies using
Property &
Construction
14%
Others
10%Manufacture/In
dustry
39%
Trade
31%
Agriculture
6%
Medium
Firms
42%
Small Firms
16% Large Firms
42%
Figure 7. Characteristic of Firm Sample
33
pre-crisis data, the banks are the main sources of funds or at least 40% of
firms‟ source of financing.
Firms using internal funds as the main sources consider funds from
head/business group (46%) and retained earnings (44%) as the main
sources. The incomes from deposit interest and foreign exchange profit
are only around 4%. Referring to credit crunch survey, the main reasons of
using internal fund are the relatively high loan rate, under utilized of their
own capital, tightness of credit procedures and the existence of banks
credit rationing.
Firms using bank loans as main source of financing come from
manufacturing sector 37,9%, while trade and property/construction have
the same portion about 20,7%, and agriculture sector only 13,8%.
Classified according to business scale, large firm 55,2%, medium firm 41,4%
and small firm only 3,4%. Small portion for agriculture sector and small
scale business because the respondent from those categories
Figure 10. Banks Loan as the main source of external fund
Figure 8. Sources of Funds Figure 9. Sources of Internal Funds
34
experiencing difficulty to obtain credit bank. Obstacles in obtaining bank
credit are tightness of collateral condition, declining cash flow, and credit
rationing.
6.2. Lending behaviour after a monetary shock
The necessary condition of the existence of bank lending channel is
whether or not a monetary policy influences the loans supply. The survey
indicates that in the case of tight money, the majority of banks (77%) will
reduce their loan supply and 23% of banks will not. As indicated by the
quantitative study, the foreign and joint-venture banks are less influenced
by the tight money than their domestic counterparts. The survey suggests
that 50% of foreign and joint-venture banks will reduce their loans in the
aftermath of the tight money policy. Meanwhile, all private non-foreign
exchange banks and regional banks reduce their lending supply. This
supports previous empirical findings (e.g. Agung, 1998) that small banks‟
reliance on the deposits as the source of funds makes their lending is more
sensitive to a monetary tightening. By contrast, foreign banks and larger
banks such as state banks and private foreign exchange banks that have
access to non-deposit funds (e.g. foreign funds) are able to shield their
lending supply from the shock. Furthermore, the banks‟ holdings of
securities enable them to protect their lending, at least in the short run.
0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 40.0%
State banks
FX Private domestic banks
Non-FX Private domestic banks
Regional banks
Foreign banks
Yes No
Figure 11. Reducing loan supply in the aftermath of monetary
shock?
35
In the case of monetary tightening reflected in an increase in the
SBI rate, banks reduce bank lending supply either by price mechanism,
through increasing the loan rate or tightening the credit conditions, and
non-price mechanism, through reducing new loans. Majority of banks
(71%) raise the loan rate in the aftermath of tight money and around
21.4% of banks reduce the loan supply. A more interesting result is that
banks that reduce the lending by rationing credit rather than by raising
the loan interest rates coming from the private banks and regional banks.
Meanwhile, state banks and foreign banks raising the interest rate in order
to reduce loan. The similar result is found in the case of monetary
loosening (a fall in SBI rate), that is, around 72% of banks reduce loan rates
and around 20% of banks raise the lending supply.
Table 8. Effects of monetary policy shock
Reduce in SBI rate Rise in SBI rate
Banks Reducing
loan
rates
Raising
new
loans Others Total
Raising
loan rate
Reducing
new
loans Other Total
State banks 2 1 1 4 3 1 1 5
50.0% 25.0% 25.0% 100.0% 60.0% 20.0% 20.0% 100.0%
Private FX banks 8 3 0 11 6 5 0 11
72.7% 27.3% 0.0% 100.0% 54.5% 45.5% 100.0%
Private Non-FX
banks 1 1 0 2 2 0 0 2
50.0% 50.0% 0.0% 100.0% 100.0% 100.0%
Regional banks 0 1 0 1 0 1 0 1
0.0% 100.0% 0.0% 100.0% 0.0% 100.0% 100.0%
Foreign and joint-
venture 7 0 1 8 9 0 0 9
87.5% 0.0% 12.5% 100.0% 100.0% 0.0% 0.0% 100.0%
Total 18 5 2 25 20 6 1 28
72.0% 20.0% 8.0% 100.0% 71.4% 21.4% 3.6% 100.0%
36
6.3. Lag in responses to a monetary tightening
As found previously that some banks do not reduce the loan supply
after a monetary tightening. This confirms the quantitative results that
suggest that the response of bank lending to a tightening of monetary
policy needs time lag. There are three possible reasons for the apparently
insensitive lending supply: (1) to maintain their relationship with their
borrowers; (2) to honour the loan commitments that have been made; (3)
to prevent the borrower‟s financial problem if the banks discontinue the
lending supply. The survey indicates that the main reason is to maintain
their relationship with borrowers and to honour loan commitments.
To finance the lending activities in the case of tight money, the
majority of banks liquidate their SBIs. This supports the empirical findings
that the banks‟ securities fall in the aftermath of the monetary tightening.
The second resort is borrowing from interbank market and selling their
bonds holdings.
0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00%
Borrowing from interbank market
Borrowing from abroad
Selling SBI
Selling bonds
Figure 12. The reasons of not to reducing the lending suply after a shock
Figure 13. Source of funds to finance lendings after a monetary shock
37
6. Summary and Conclusions
We have investigated the existence of bank lending channel of
monetary transmission in Indonesia before and after crisis. Given existence
of „bank dependent borrowers‟ as the secondary condition of bank
lending channel clearly satisfied, our study particularly focuses on the first
condition of the bank lending channel to exist; that is, whether a
monetary policy affects the quantity of bank lending. We use three
different methods to achieve robust conclusions. First, using Bernanke-
Blinder type of VAR we examine responses of banks‟ balance sheet
(deposits, lending and securities holdings) to a monetary shock measured
by SBI rate and interbank rates. Second, using the restricted version of
VECM to identify supply of and demand for credit and examine whether
the short run adjustment toward equilibrium is dominated by the supply
determined as suggested by the credit channel hypothesis. Third, we use
bank-level panel data to investigate in detail differential behaviour of
bank lending, especially with regard to their capital strength and asset
size.
Aggregate evidence show that a monetary policy is able to affect
bank lending with a lag due to ability of banks to insulate the decrease in
deposits by liquidating their securities holdings. This is conducted by bank
to serve the commitment loans that have been made prior to the
monetary shock. Empirical results with disaggregate data across bank
categories indicate that after a monetary shock, in particular in the period
of post crisis, there is a flight to quality of deposits especially from private
domestic banks to foreign banks and state banks. Accordingly, lending
of these categories of banks is less sensitive to a monetary shock
compared with that of private banks.
38
A disaggregation of total bank loans into corporate and individual
lending demonstrates that the response of aggregate lending to firms is
less sensitive to a monetary policy. By contrast, the loans for individuals
drop significantly in the aftermath of monetary shock. This may be
explained by what so-called the „flight to quality‟ phenomenon. That is, in
a monetary contraction, to compensate the decline in cash flow, the
creditworthy borrowers have access to short-term loans, while loans to the
less creditworthy borrowers, such as individuals, will be rationed. .
Disaggregation of banks according to their capital strength, we found
that the effect of monetary policy on bank lending is stronger for banks
with low capital. From time series and panel data estimations, the study
found that efficacy of a monetary policy in influencing the bank lending
and thus investment is stronger in the aftermath of the crisis, especially in
the case of monetary contraction. Ineffectiveness of monetary policy in
affecting the bank lending prior to the crisis was due to banks‟ ability to
access funds from international sources. In the wake of the crisis, given
deterioration of bank capital and high credit risk, an increase in interest
rate as a result of a monetary tightening raises the probability of loan
default, hence banks become reluctant to extend credits. This findings
lend support the existence of asymmetric effect of monetary policy;
stronger in the recession than in the boom periods.
39
References
Agung, Juda (1998). Financial deregulation and bank lending channel of
monetary policy in developing countries: The case of Indonesia, Asian
Economic Journal, Vol. 12, No.3, 273-294.
Agung, J., Bambang Kusmiarso, Bambang Pramono, Erwin G Hutapea,
Andry Pasmuko, Nugroho J Prastowo (2001). Credit Crunch in Indonesia in
the Aftermath of the Crisis: Facts, Causes and Policy Implications. Bank
Indonesia, Directorate of Economic Research and Monetary Policy.
Bernanke, B., 1993a, Credit in the macroeconomy. Fed Res Bank of New
York Quarterly Review, Spring 1993, pp. 50-70.
Bernanke, B., 1993b, How important is the credit channel in the
transmission of monetary policy? A comment, Carnegie-Rochester Series
on Public Policy, December 1993, pp. 47-52.
Bernanke, B., 1995, What do we know about how monetary policy affects
the economy, Fed Res Bank of St Louis Economics Review, May/June
1995, pp. 127-42.
Bernanke, B. and Blinder, A.S., 1988, Credit, Money and Aggregate
Demand, The American Economic Review , June 1983, pp. 257-76.
Bernanke, B. and Blinder, A.S., 1992, The federal funds rate and the
channel of monetary transmission, The American Economic Review
September 1992, pp. 901-21.
Bernanke, B. and Gertler, M., 1989, Agency cost, net worth and business
fluctuations,
The American Economic Review March 1989, pp. 14-31.
Bernanke, B and Gertler, M., 1995, Inside the black box: credit channel of
monetary transmission mechanism, Journal of Economic Perspective, Fall
1995, pp. 27-48.
Bernanke, B., Gertler M., and Gilchrist, S., 1996, The financial accelerator
and the flight to quality. Review of Economics and Statistics February
1996, pp. 1-15.
40
Callen, T. and Reynolds, P., 1996, Capital market development and
financial deregulation in Thailand and Malaysia: implications for financial
structure and the monetary transmission mechanism, Paper presented in
the Conference on Macroeconomic Issues Facing ASEAN Countries,
Jakarta.
Calomiris, C.W and Hubard, G., 1990, Firm heterogeneity, internal finance
and credit rationing. Economic Journal, 100, pp. 90-104.
Cecchetti, S.G., 1995, Distinguishing theories of the monetary transmission
mechanism, Fed Res Bank of St Louis Economic Review, May/June 1995,
pp. 83-100.
Cole, D. and Slade, B.S., 1990, Indonesian Financial Development: A
different sequencing? Harvard Institute for International Development
HIID, Harvard University, Cambridge, Massachusetts unpublished.
Cole, D. and Slade, B.S., 1996, Building a modern financial system: The
Indonesian experience, Cambridge University Press.
Dale, S and Haldane, H., 1994, Interest rates and the channel of
monetary transmission: Some sectoral estimates, European Economic
Review 39, pp. 1611-25.
Enders, W., 1995, Applied Econometric Time Series, John Wiley and Sons,
Inc.
Gertler, M and Gilchrist, S., 1993, The role of credit market imperfections in
the monetary trasmission mechanism: Arguments and evidence,
Scandinavian Journal of Economics, 951, pp. 43-64.
Gertler, M and Gilchrist, S., 1994, Monetary policy, business cycles and the
behaviour of small manufacturing firms, Quarterly Journal of Economics,
May 1994, pp. 309-40.
Himmelberg, C.P., and Morgan, D.P., 1995, Is bank lending special?, in
Peek, J., and Rosengren, E.S. (eds), Is bank lending important for the
transmission of monetary policy?, Federal Reserve Bank of Boston,
Conference Series No. 39, June 1995.
International Monetary Fund, 1995, Capital Flows to the Asia Pacific
Region, Occasional Paper No. 122. March 1995.
41
Kakes, J. (2000). Monetary Transmission in Europe. The Role of Financial
Markets and Credit. Edward Elgar: Cheltenham.
Kashyap, A.K. and Stein, J.C., 1995, Impact of monetary policy on bank
balance sheets. Carnegie-Rochester Series on Public Policy, June 1995,
pp. 151-95.
Kashyap, A.K. and Stein, J.C., (2000), What do a million banks have to say
about the transmission of monetary policy?, American Economic Review.
Peek, J and Rosengren, E. ,1995, The capital crunch: neither a borrower
nor a lender be, Journal of Money Credit and Banking, August 1995, pp.
625-38.
Ramey, V, 1993, How important is the credit channel in the transmission of
monetary policy?, Carnegie-Rochester Conference Series on Public
Policy, Dec 1993, pp. 1-45.
Romer, C.D. and Romer,D. ,1990, New evidence on the monetary
transmission mechanism, Brooking Papers on Economic Activity 1990:1,
pp. 149-213.
Spiegel, M, 1995, Sterilisation of capital inflows through the banking
sector: evidence from Asia, Federal Reserve Bank of San Fransisco
Economic Review, Vol 1995, No. 3, pp. 17-34.
42
Appendix A. Data: sources and definition
Data are monthly from January 1991 - December 2000.
1. Macroeconomic data:
SBI interest rate: It is the end-of-period 1-month SBI rates published
in Weekly Report of Bank Indonesia.
Money market (PUAB) interest rate: It is the end-of-period 1-month
interbank call money rates published in Weekly Report of Bank
Indonesia.
Real GDP: Monthly data of the real GDP is interpolated from
quarterly real GDP published in Indonesian Financial Statistics, Bank
Indonesia. The interpolation was performed by the piece-wise
cubic spline method. For early periods when the quarterly data
were not published officially, the data were obtained directly from
Indonesian Central Bureau of Statistics.
Production Index: Monthly data of the production index is
interpolated from quarterly production index published by Central
Bureau of Statistics. The interpolation was performed by the piece-
wise cubic spline method.
Prices: Consumer price index published in the Indonesian Financial
Statistics, Bank Indonesia.
43
2. Banks’ balance sheet data:
Banks‟ balance sheet data is obtained from Monthly Commercial Bank
Report.
Deposits: Consist of demand deposits, savings deposits and time
deposits both in Rupiah and foreign currency, excluded certificate
deposits.
Total loans: total loans extended by commercial banks both in
Rupiah and foreign currency.
Working capital loans: loans extended for firms‟ working capital
both in Rupiah and foreign currency.
Investment loans: loans extended for firms‟ investment both in
Rupiah and foreign currency.
Loans to individuals: loans extended to households mainly for
durable goods, real estate and credit cards.
Loans to firms: loans extended to private enterprises, in the form of
either working capital or investment loans.
43
Appendix B. Results for cointegration tests
AGGREGATE
1991:01 - 2000:12 1991:01 - 1997:07
RPUAB RSBI RPUAB RSBI
Rank p Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat
1 137,63** 57,42** 136,93** 54,60** 155,40** 58,02** 182,58** 67,75**
2 80,21** 34,84* 82,33** 38,13* 97,38** 35,89* 114,84** 44,23**
3 45,37 23,36 44,20 19,07 61,49** 27,63* 70,61** 36,50**
4 22,00 11,90 25,13 17,53 33,86* 19,67 34,11* 18,43
5 10,11 10,01 7,60 6,67 14,19 9,88 15,68* 11,08
6 0,09 0,09 0,93 0,93 4,31* 4,31* 4,59* 4,59*
STATE BANKS
1991:01 - 2000:12 1991:01 - 1997:07
RPUAB RSBI RPUAB RSBI
Rank p Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat
1 165,96** 65,94** 183,98** 67,45** 109,83** 44,24* 128,99** 49,69**
2 100,02** 39,73** 116,54** 51,09** 65,59 23,35 79,30** 35,15*
3 60,29** 30,35* 65,45** 35,45** 42,24 21,12 44,15 20,94
4 29,93* 21,96* 30,00* 23,85* 21,12 11,54 23,21 13,57
5 7,98 7,92 6,16 5,27 9,58 5,82 9,64 6,31
6 0,057 0,06 0,88 0,88 3,76* 3,76* 3,33 3,33
44
PRIVATE BANKS
1991:01 - 2000:12 1991:01 - 1997:07
RPUAB RSBI1M RPUAB RSBI1M
Rank p Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat
1 166,30** 84,91** 153,93** 64,18** 154,09** 52,83** 154,55** 48,76**
2 81,39** 36,11* 89,75** 59,07** 101,26** 39,84** 105,79** 40,25**
3 45,28 22,43 30,68 18,49 61,42** 27,77* 65,53** 34,44**
4 22,86 14,89 12,19 6,99 33,65* 22,15* 31,10* 17,99
5 7,96 7,93 5,19 5,11 11,50 8,34 13,11 7,56
6 0,03 0,03 0,08 0,08 3,16 3,16 5,54* 5,54*
FX-BANKS
1991:01 - 2000:12 1991:01 - 1997:07
RPUAB RSBI1M RPUAB RSBI1M
Rank p Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat
1 163,25** 82,01** 147,51** 70,14** 146,21** 45,17** 147,78** 41,25*
2 81,24** 33,92* 77,37** 49,14** 101,04** 36,19* 106,53** 37,80*
3 47,32* 24,33 28,23 13,38 64,86** 28,07* 68,73** 32,34**
4 22,98 16,39 14,85 9,66 36,79** 23,05* 36,39** 21,37*
5 6,59 6,52 5,19 4,91 13,73 11,69 15,02 11,43
6 0,07 0,07 0,28 0,28 2,05 2,05 3,59 3,59
45
FOREIGN BANKS
1991:01 - 2000:12 1991:01 - 1997:07
RPUAB RSBI1M RPUAB RSBI1M
Rank p Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat
1 122,64** 54,41** 131,23** 57,07** 148,73** 64,96** 147,66** 64,74**
2 68,23 29,46 74,16* 37,53* 83,77** 37,21* 82,91** 37,00*
3 38,77 17,93 36,64 18,21 46,56 27,83* 45,91 25,94
4 20,84 14,42 18,42 12,10 18,73 12,66 19,97 11,73
5 6,42 6,28 6,33 5,46 6,07 4,83 8,24 8,04
6 0,14 0,14 0,87 0,87 1,24 1,24 0,19 0,19
LOAN TO INDIVIDUALS
1991:01 - 2000:12 1991:01 - 1997:07
RPUAB RSBI1M RPUAB RSBI1M
Rank p Trace Stat Max-Eigen Stat Trace Stat
Max-Eigen
Stat
Trace
Stat
Max-Eigen
Stat
Trace
Stat Max-Eigen Stat
1 242,54** 110,94** 203,52** 66,96** 136,44** 50,09** 153,03** 46,44**
2 131,59** 57,24** 136,56** 63,29** 86,35** 35,77* 106,58** 41,62**
3 74,35** 38,33** 73,27** 39,70** 50,59* 25,53 64,95** 36,32**
4 36,02** 18,63 33,56* 18,69 25,05 18,31 28,63 22,10*
5 17,39* 16,83* 14,87 14,79* 6,74 6,55 6,52 6,52
6 0,57 0,57 0,07 0,07 0,19 0,19 0,00 0,00
46
LOAN TO PRIVATE ENTERPRISES
1991:01 - 2000:12 1991:01 - 1997:07
RPUAB RSBI1M RPUAB RSBI1M
Rank p Trace Stat Max-Eigen Stat Trace Stat
Max-Eigen
Stat Trace Stat Max-Eigen Stat Trace Stat
Max-Eigen
Stat
1 265,79** 109,59** 232,07** 89,59** 138,39** 42,63*
2 156,19** 76,11** 142,48** 78,46** 95,76** 40,34**
3 80,09** 50,18** 64,02** 32,36** 55,42** 27,64*
4 29,91* 20,37 31,66* 23,96* 27,78 18,06
5 9,55 9,47 0,07 7,66 9,72 8,28
6 0,08 0,08 0,00 0,05 1,44 1,44
INVESTMENT LOANS
1989:01 - 2000:12 1989:01 - 1997:07
RPUAB RSBI1M RPUAB RSBI1M
Rank p Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat
1 147,75** 54,86** 191,32** 73,13** 135,24** 59,84** 141,77** 63,61**
2 92,89** 36,37* 118,19** 43,00** 75,39* 33,33 78,15** 29,21
3 56,52** 23,77 75,19** 35,50** 42,06 19,31 48,94* 27,29*
4 32,75* 20,61 39,69** 21,27* 22,75 17,15 21,65 11,81
5 12,14 10,78 18,42* 13,26 5,59 4,65 9,84 5,65
6 1,36 1,36 5,15* 5,15* 0,94 0,94 4,19* 4,19*
47
WORKING CAPITAL LOANS
1989:01 - 2000:12 1989:01 - 1997:07
RPUAB RSBI1M RPUAB RSBI1M
Rank p Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat Trace Stat Max-Eigen Stat
1 272,01** 132,95** 222,94** 90,08** 196,10** 78,15** 190,79** 76,19**
2 139,06** 60,07** 132,86** 64,35* 117,95** 54,98** 114,59** 42,84**
3 78,98 49,89** 68,51** 36,25** 62,97** 36,31** 71,74** 39,27
4 29,09 20,63 32,26* 25,04* 26,67 16,89 32,47* 20,51
5 8,46 8,45 7,22 7,15 9,77 6,99 11,96 8,23
6 0,00 0,00 0,07 0,07 2,78 2,78 3,73 3,73